In some types of networks, such as networks compliant with the Bridged Ethernet standards, a protocol is used to configure the paths over which data with a specific destination is transmitted, i.e., to obtain information via which network nodes data (often in the form of frames or packets) has to be sent.
In Bridged Ethernet networks this protocol is called a spanning tree protocol which is used to configure a spanning tree over which data traffic is transmitted. The spanning tree provides a unique path between any two nodes in the network. As a part of the spanning tree protocol each bridge in the network uses a learning algorithm to store in an address table the directions in which received frames have to be forwarded. At times, the spanning tree may be altered by the spanning tree protocol, for example in case of failure of a link between two bridges. As part of the altering process, the address tables have to be flushed, at least partially, and rebuilt using the learning algorithm. However, the rebuilding may result in an incorrect address table. For example, if a bridge receives data transmitted both via the old spanning tree and the new spanning tree, the rebuilt address table will provide an addressing scheme based on the old, now incorrect, spanning tree which is likely to result in errors in the data traffic.
From the IEEE 802.1D standard, it is known to prevent incorrect rebuilding of the address table by blocking datalinks for rather a long time before starting the rebuilding of the address table, thereby to ensure that data frames transmitted before the altering of the spanning tree have disappeared from the network, for example because the frames reached their destination or were thrown away or discarded by nodes in the network.
However, a disadvantage of this IEEE method is that the entire network may not be used for a long time, typically around 50 seconds.